ubuntu 18.4 python3.8

[songsh]

trainer = Trainer(data_bundle.get_dataset('train'), model, loss=loss, optimizer=optimizer, File "/home/dony/.local/lib/python3.8/site-packages/fastNLP/core/trainer.py", line 500, in init self.model = _move_model_to_device(model, device=device) File "/home/dony/.local/lib/python3.8/site-packages/fastNLP/core/utils.py", line 424, in _move_model_to_device

assert torch.cuda.device_count() > device, "Only has {} gpus, cannot use device {}.".format( AssertionError: Only has 0 gpus, cannot use device 0.

[yhcc]

就是说没有检测到可用gpu，但是却申明要用gpu。把trainer的device修改为None可以规避这个问题。如果您的电脑确实有gpu，这可能是(1) 显卡驱动没安装好, 可以通过nvidia-smi检查; (2) pytorch装成了cpu版本.